Fix code caching for OpenMP regions (CPU and GPU)

1) Fix dispatcher pickling to only return output_types if it is a object mode lifted with. (#29)
2) Make openmp_tag pickleable by not storing the arg if it contains a llvmlite AllocaInstr.
3) Don't pickle anything for LiftedCode's function IR except the arg count and the func_id.
4) Return all of openmp_region_start for pickling.
5) Comment on why we need new class.

Author: DrTodd13

numba/core/consts.py

@@ -21,6 +21,9 @@ def __init__(self, func_ir):
         self._func_ir = weakref.proxy(func_ir)
         self._cache = {}
 
+    def __repr__(self):
+        return str(self._cache)
+
     def infer_constant(self, name, loc=None):
         """
         Infer a constant value for the given variable *name*.

numba/core/dispatcher.py

@@ -1101,6 +1101,16 @@ def __init__(self, func_ir, typingctx, targetctx, flags, locals):
                                  can_fallback=True,
                                  exact_match_required=False)
 
+    class LCFuncIR:
+        """ This is a minimalistic reimplementation of Numba's regular FuncIR
+            class.  We only need these two parts for caching.  We either don't
+            have the information to fill in the regular FuncIR or there are
+            problems if we do.
+        """
+        def __init__(self, arg_count, func_id):
+            self.arg_count = arg_count
+            self.func_id = func_id
+
     def _reduce_states(self):
         """
         Reduce the instance for pickling.  This will serialize
@@ -1110,8 +1120,8 @@ def _reduce_states(self):
         NOTE: part of ReduceMixin protocol
         """
         return dict(
-            uuid=self._uuid, func_ir=self.func_ir, flags=self.flags,
-            locals=self.locals, extras=self._reduce_extras(),
+            uuid=self._uuid, func_ir=self.LCFuncIR(self.func_ir.arg_count, self.func_ir.func_id),
+            flags=self.flags, locals=self.locals, extras=self._reduce_extras(),
         )
 
     def _reduce_extras(self):
@@ -1227,7 +1237,10 @@ class LiftedWith(LiftedCode):
     can_cache = True
 
     def _reduce_extras(self):
-        return dict(output_types=self.output_types)
+        if hasattr(self, "output_types"):
+            return dict(output_types=self.output_types)
+        else:
+            return dict()
 
     @property
     def _numba_type_(self):

numba/openmp.py

@@ -52,6 +52,7 @@
 
 llvm_binpath=None
 llvm_libpath=None
+
 def _init():
     global llvm_binpath
     global llvm_libpath
@@ -80,6 +81,7 @@ def _init():
 
 _init()
 
+
 #----------------------------------------------------------------------------------------------
 
 class NameSlice:
@@ -166,6 +168,17 @@ def __init__(self, name, arg=None, load=False, non_arg=False, omp_slice=None):
         self.non_arg = non_arg
         self.omp_slice = omp_slice
 
+    def __getstate__(self):
+        state = self.__dict__.copy()
+        if isinstance(self.arg, lir.instructions.AllocaInstr):
+            del state['arg']
+        return state
+
+    def __setstate__(self, state):
+        self.__dict__.update(state)
+        if not hasattr(self, "arg"):
+            self.arg = None
+
     def var_in(self, var):
         assert isinstance(var, str)
 
@@ -1120,6 +1133,13 @@ def __init__(self, tags, region_number, loc, firstprivate_dead_after=None):
         self.alloca_queue = []
         self.end_region = None
 
+    def __getstate__(self):
+        state = self.__dict__.copy()
+        return state
+
+    def __setstate__(self, state):
+        self.__dict__.update(state)
+
     def replace_var_names(self, namedict):
         for i in range(len(self.tags)):
             if isinstance(self.tags[i].arg, ir.Var):